ietf-smtp
[Top] [All Lists]

Re: Minor is. It's not a pardigm change

2008-03-31 14:49:54

On Mon, Mar 31, 2008 at 11:56:51AM -0400, Hector Santos wrote:

But I think overall, there are two discussions going on here.

  1) A vs AAAA fall backs, and

1a: "A" fallback for IPv4 and none for IPv6
1b: "A" fallback for IPv4 and "AAAA" fallback for IPv6
1c: "A" fallback for all
1d: "A" and "AAAA" fallback for all
1e: "A" or "AAAA" fallback for all

I think 1e would be insane. Choice is not good.
1c wouldn't work for IPv6-only hosts
1d wouldn't work for IPv4-only hosts, plus you do not want to
change existing installations.

Which leaves 1a and 1b.  And I do think that most of us argue
for either 1a or 1b.  I am not in favour of it, but I can see
why people would want 1b.

Getting rid of the fallback has nothing to do with:

  2) Mandating MX for Email Security Policies

I think this one can stop; it is clear to most that this is not
the right time and place.



But there are other reasons why fallback should go.  IPv6 is
designed so that everybody can have all the addresses they want,
each one can have a name.

Back then, when computers were huge and expensive, it made sense
that every host would have a hostname which would also work as an
email domain name.

Now and in the near future, this is quite different. I do not want
my microwave oven, my thermometers, my light switches, my vacuum
cleaning robot, the fridge, and whatever other appliance with an
address, to be seen as mail domains.

You really do not want to publish something like "MX 0 ." for
each and every one of them, except for the one or two hosts that
do receive mail.

If the sending side does not find an MX record, should it really
continue to query for an AAAA record and/or an A record?

Yes, for IPv4, because this is the standard. Back then it was a
good thing to do, but nowadays it is no longer necessary (only
conveniant for some). Nevertheless: don't change it.

No, for IPv6, because right now is the correct time and place to
get rid of it.


I do appreciate that RFC 974 mentions hostname, not A record, and
thus implicitly also mentions AAAA records. But that RFC also
discusses WKS queries. Things change.

I think it would be a huge mistake to have "AAAA" fallback just
because it is similar to "A" fallback.